TransNILM: A Transformer-based Deep Learning Model for Non-intrusive Load Monitoring

Non-intrusive load monitoring (NILM) is crucial for energy efficiency in demand-side management. The conventional methods, such as combination optimization or the factorial hidden Markov model, are not effective to obtain the energy consumption information of each appliance. To overcome the limitations of these conventional methods, deep learning-based methods are widely developed. Inspired by recent advances in deep learning models, this paper proposes a novel transformer-based deep learning model for NILM. In this proposed model, transformer-layer enhanced convolutional neural networks (CNN) are firstly utilized for feature learning, a temporal scaling module is then developed for multi-scale information learning. Finally, these learned features are fed into a decoder which consists of three residual GRU modules and a transformer-based CNN module. To evaluate the proposed method, this paper conducts experiments on one residential dataset and one commercial dataset and compares our proposed model with several state-of-the-art methods. The experimental results demonstrate that the proposed method achieves better performance and has good generalization capability.